Industrial Control Systems (ICSs) are widely applied in critical infra-structures to provide essential utilities to society. As a result, protecting them against terrorist activities, defects and cyber-attacks are crucial. The tremendous rise in cyber-attacks against Industrial Control Systems has triggered the advancement of security systems for accurate and timely detection of ICS cyber-attacks. In this paper, machine learning based ensemble XG-Boost algorithm for anomaly detection in industrial control systems proposed to identify cyber-attacks in ICS environments. Intrusion detection system based on anomalies is more efficient than other forms of intrusion detection system. Proposed system is a build with multiple machine learning algorithms, on Hardware-in-the-Loop (HIL) industrial process dataset (HAI). The HAI dataset was gathered from a realistic industrial control system testbed augmented with a hardware in the loop simulator that simulated pumped storage hydropower generation and steam-turbine power generation. XG-ADICS consist of 4 main steps, starting from data pre-processing then model training followed by model testing and at last classification of anomalous or normal observation. System findings demonstrate that by employing XG-ADICS, we can attain greater ac-curacy 99.96 % with a higher F1 score of 0.9986. The results suggest that the XG-Boost algorithm outperforms other algorithms in finding abnormalities. results demonstrates that the suggested system is appropriate for an ICS environment

yber Physical Systems, Cyber Physical Security, Industrial Control System, Machine Learning, Intrusion Detection System.